Automatic structure and fault detection of semiconductor micrograph images

grantor: University of TorontoAs with almost all materials, semiconductors exhibit periodicity in their crystal lattice, and the surface pattern evident a the surface provides researchers with information that is useful in many experiments. Furthermore, ' Scanning Probe Microscopy', facilitates the acquisition of ' atomic resolution' images, and provides insight into the atomic lattice pattern of sample being investigated. In addition to the aforementioned lattice pattern, scanning probe microscopy techniques are capable of imaging surface impurities such as foreign atoms, vacancies or other surface defects. Yet, manually performing an analysis of semiconductor micrographs is a time consuming and boring task. Since computers are perfectly suited to perform the repetitive and tedious jobs, it is evident that a hand analysis of a micrographs is essentially a 'waste' of time. In this thesis, a high-level knowledge-band computational technique is developed which extracts the relevant elements from a semiconductor micrograph, pinpoint surface faults, and delineate the lattice pattern at the surface. (111)7x7 Silicon in the material upon which focus is placed because of its ubiquity, simplicity, and simple pattern. The system applies image processing methods such as mathematical morphology, and roundness measurement of recursively A spatial analysis which uses apriori knowledge of the makeup of (111)7x7 silicon, is then used to find parameters of the surface pattern. The results provided by this system for structure extraction are very good, with an average structural detection rate near 90%. In contrast, fault detection does not prove to be as effective giving adequate but sometimes faulty and ambiguous results. Overall, the system proves to provide a great savings in time over hand-analysis, in addition to increasing the informational content of the micrograph through the overlaying of data related to the material's surface structure.M.A.Sc